how come welding is so tough to automate (how come there’s lots of welders, but factory work was automated)
In: Engineering
They have robotic welders. But the part that’s hard to automate is that a lot of welding jobs have unique circumstances and aren’t repeated. So taking time to setup and reprogram a robot for a one time job is more costly and time consuming than using a human laborer (for now).
Robotic welders are best for jobs where you have to weld the same thing the same way dozens/hundreds/thousands of times.
Most welding is automated.
But you have the same issues as with any production: using expensive tooling/robots is only efficient if you are producing a large number of items with the same dimensions.
So random T metal bars and shit that are mass produced will be welded with machines/robots.
But if you want to weld together stuff on a building site? A trained human is muuuuch cheaper and faster than setting up a welding robot to find its way around in 3d space.
Welding that happens in factories is automated to a large degree. Only welding in factories that isn’t is when you need to frequently weld different things, so setting up a welding robot makes no sense, or when the shapes are kinda convoluted where it is again cheaper to use the humans brain than trying to make an AI robot welder.
The challenge is being able to adapt to different types of jobs. Lots of robotic welding is used in many industries, but usually for performing the identical weld on thousands (to millions) of identical joints.
I would not be surprised, however, if adaptive expert AI systems start being able to assess the type of job they are given and change their technique appropriately. But this will have to happen *very carefully* as failed welds can kill people.
I’ve done robotic welding. It’s OK if everything in the process before robotic welding is handled by robots with tight tolerances. If there’s human involvement prior to that then it might fall outside tolerance. Once the robot hits something outside it’s programming it could potentially mess everything up after.
I did robotic welding on toolboxes for truck beds. If the fit wasn’t perfect, welds ran off track, machine got gunked up, tips get burnt up and all kinds of stuff. I had to step in and do some light reprogramming work or finish welding and do touch up stuff.
They also take up significant real estate in the building they are in. There’s also the flexibility of a human VS robot. Every time a new part needs done there’s retooling, programming, tear down and set up for configuration. Where as I can make adjustments on the fly and weld different parts up in a few minutes with minimal down time.
Long and short of it is they are good for repeated process in a significant amount. But short run stuff it’s really not feasible for robotic work.
There are lots of different things which all work together when welding. You have to be very careful about how much heat you put into the part, too little and you won’t melt the metal together, too much and you’ll burn through or warp the part into the wrong shape, or screw up the metal’s properties.
To control all of that you can play with the arc current, arc gap, travel speed, shield gasses, filler metals, electrodes, and all of those will have a different effect depending on the type of metal or alloy you have.
In short, it’s a very skilled job with a lot that can go wrong, and if you only have one part which is a weird shape like a building strut or an expensive component you want to make sure you get it right. Humans are very good at thinking of lots of variables together at the same time so if the weld starts to go wrong, they can feel how to increase power or back out to keep it on the level. If it’s the same short weld over and over again like a car factory it’s easy to dial in the parameters to automate it.
There’s also some limitations with robots, for example if you have 2 metal surfaces put together in a butt joint they can both be shiny and cameras will have difficulty doing accurate seam tracking without expensive scanning and rendering. So it’ll work if the robot knows exactly where to find the joint, but can be difficult if you show the robot something for the first time. Usually cheaper to hire a person than to invest in all the equipment.
Worth saying this is only the case for ‘traditional’ arc welding like MIG or TIG welding, processes like laser or electron beam welding are usually 100% automated. One of the big problems in the industry now is we are hugely short of engineers who know how to program laser welders properly.
Any one weld is very easy to automat, the issue is there is hundreds of welds on all different types of things so no one machine can do all of them except a person
I’m a welder and I’ve worked in automated welding shops. The robots constantly fucked up so guys like me were on site to fix what the bot couldn’t do. If parts started coming out wrong or unfinished, they’d stop the line, make adjustments, and run her again until things started to go sideways. Meanwhile me and a few other guys would fix what was wrong with the parts and the company would lose less money.
You would be surprised how much welding is automated. For example, the majority of joints in pipelines are automated ( in position with orbital ) . Obviously automotive industry is an example of all robotic. Heavy fabrication is heavily reliant to robotics as well.
Big change I have seen ( I work as automation specialist in Welding ) is shipyard construction. It’s simply not true that a robot needs prefect fit-up and programming. Quite the opposite now where systems will via vison system and / or the CAD drawings will do the path programming on the fly. You can take a micropanel with all the stiffeners and bulkheads and the robotic systems will go in between bulkheads and through the vision will program all the out of position and flat position welds and do them far more efficiently than a welder
I also spend much of my time in the wind energy and offshore sector. All the heavy welding on the monopiles, towers etc are done with automation with vision and laser seam tracking. The one part that is always tough with automation is the TKY nodes as well as ladder mounts and other dicky fitup . Having said that all the fabricators are make great trides the past couple of years to even automate these welds.
Bottom line is adaptive vision has really come along in the past few years . Current trend is beyond that where it’s not only adaptive but using artificial intelligence it will make decision on how to adapt the welding path and parameters.
Latest Answers